Discussion 311 Group 4

Introduction

During the research and distribution periods of the COVID-19 vaccines, social media-driven anti-vaccination campaigns have gained popularity, particularly among the Facebook1 and Instagram2 communities. We want to find whether there is a relationship between environmental and societal factors and vaccine adoption rates.

We theorize that more urban and educated populations are more likely to have higher rates of vaccine adoption than those populations with a higher proportion of rural residents and less education, and higher median age also contribute to the higher vaccination rate.

Background

Variables

Vaccine Data

  • one_dose_at_least_p: Percent people get at least one dose
  • completed_series: People who completed the series
  • completed_series_p: Percent people who completed the series

Urban and Rural Data

  • population: Total population of the county (All ages)
  • urban: Population living in an urban environment of the county
  • rural: Population living in a rural environment of the county

Education Data

  • pct_m_hs: Percent males estimated high school graduate or higher
  • pct_f_hs: Percent females estimated high school graduate or higher
  • pct_m_bach: Percent males estimated bachelor’s degree or higher
  • pct_f_bach: Percent females estimated Percent bachelor’s degree or higher

Median Age data

  • median_age: Median age in each county
  • county: County in Wisconsin

Note for Missing Data:

  • Although the education data is only for 25+, we estimate that the vaccination rate of those under 25 is insignificant and thus estimate the education population groups on the urban and rural population data.

  • There are some missing value in wi_zip.csv, wi_income.csv, wi_age.csv, and we dropped them for later process.

Way of collection

Vaccine Data

Source: Wisconsin Department of Health Services & The Wisconsin Immunization Registry (WIR)3

Every night at 11:30 pm we extract vaccine administration data from WIR that will be reported on the DHS website by 2:00 pm the following day. WIR is a live system and providers are constantly sending immunization data. Therefore, data will look different if it is extracted at a different time of day.

Vaccination administration: The cumulative number of COVID-19 vaccines administered. The Vaccine Distribution Summary includes all vaccine doses administered by Wisconsin vaccine providers. This includes doses administered to people who resided out-of-state, but who live, work, or study in Wisconsin and qualify for vaccination in-state. This provides information to track the allocation, distribution, and administration of vaccine by Wisconsin’s vaccinators. The COVID-19 Vaccines for Wisconsin Residents dashboard displays data for Wisconsin recipients of the vaccine. This information is used to inform vaccination coverage for the state.

Vaccine dose: One vaccine dose is one vaccine product (like a shot or a nasal spray). Some vaccines require two or more doses to protect you fully against a disease. Other vaccines give you enough protection to fight the disease after just one dose or shot.

Series completion: Many vaccines require multiple doses spaced out by weeks, months, or years to provide the best protection against a disease. Once someone receives the recommended number of doses within the correct timeframe, their series is considered complete.

Urban/Rural, Income, and Education Data

These datasets are from previous class assignments, particularly Homework 64.

Zip Code and County Information

The data were from ZipCodesToGo, combined with Longitude/Latitude Coordinates for visual maps5.

Wisconsin Median Age

The data was from a local news report which was based on the U.S. Census Bureau6.

Note: Vaccine Data and Wisconsin Median Age are from government, so they are valid and legitimate. Others are from website and assignments, so they may not be so trustful.

Unusual factors that could affect results

The data of education and income are from the assignment, they may not to so up-to-date compared to the vaccination rate data. Some data obtained through unofficial websites may not be do trustworthy.

Backgroud description & relation to topic

People with different levels of education may have different understandings of vaccines. One theory is that populations with limited education may think of vaccines, especially newly developed vaccines as particularly harmful or not well-studied; whereas, populations with more education might better understand the rigorous safety standards and trials the vaccine has been scrutinized by.

Concerning the urbanization of the environment, differences in residential densities could affect accessibility and/or the speed of distribution.

At the end of the study, we can have a broad understanding of the demographic of vaccine recipients.

Analysis

Vaccination Visualization Across WI

Urbanization Rate

Note: The red dot we marked is the position of the capital Madison.

Places that are more urbanized such as Madison, Milwaukee, and Green Bay, tend to have a higher rate of vaccinations.

Among the urban cities, Madison has the lightest color, which means the vaccination rate in Madison is the highest. One possible reason to account for this is that the campus of the University of Wisconsin-Madison is at Madison, therefore, the education level of people there is also higher. Thus, the vaccination rate there would be higher.

Interestingly, there are smaller counties with less urbanization rate (e.g. Door and Bayfield, which are small yellow dots in the graph) that are particularly highly vaccinated versus other, larger counties, so that might be other factors cause such higher vaccination rate.

Level of Education

Places that are more urbanized tend to have a higher level of education.

Among communities with a higher mean education level, the vaccination rate tends to be higher. This can also explain why the places like Sister Bay and Moquah (yellow dots in northern and eastern area) are with high vaccination rate, despite being in rural counties.

Age

As we analyzed the data between vaccination rate and urban percent, education level, we found out that some of the locations may have a conflict with our assertion. Places such as Door and Bayfield (yellow-green but relative small spots in first and second map graph above) have an abnormally high vaccination rate where their education level and percent urban are not as high as big cities. We recognized that different counties have different age group distribution. Especially, for Door and Bayfield, they are particular high number of retired people. We made an assumption that the high average age in these counties maybe the reason causing it to have a high vaccination rate.

From the graph, these counties did show that a high median age than other countries. As we did a further investigation, we found out that the government’s policy of distribution vaccine has relation with the age. According to the CDC’s COVID-19 Vaccine Rollout Recommendations, there are three phases of distributing vaccines. The second phase who are eligible to get the vaccine are people aged 75 years and older where the time that the data was collected. This might be a direct reason why these places have an abnormally high vaccination rate.

Summary for this part: Places that are more urbanized tend to have a higher rate of vaccinations. Among communities with a higher mean education level, the vaccination rate tends to be higher. Places with more elder people tend to have higher vaccination rate.

Education Level and Vaccination

In the above graph, we performed a regression on how the vaccine rate and education level are correlated with each other. For both genders, it seems that there is a strong and positive linear relationship.

We will then do a residual plot to prove whether this correlation is really linear.

Residual analyze:

But the residual plot does not resemble random scatter around the horizontal line. So it suggests that the relationship between Age and Length is not linear.

In the above graph, we performed a regression on how the vaccine rate and education level are correlated with each other. For both genders, it seems that there is a strong and positive linear relationship.

We will then do a residual plot to prove whether this correlation is really linear.

Residual analyze:

But the residual plot does not resemble random scatter around the horizontal line. So it suggests that the relationship between Age and Length is not linear.

Summary for this part: There is some relationship between education rate and willingness to get vaccined, but the relation is not linear.

Income and Education Level

The two graphs above shows that with higher education level, people tend to have higher income.

We can base from this an indirect relationship that because higher education levels are associated with higher vaccine adoption rates, so too will higher levels of income.

Since the vaccinations now are free, we can not assert that there is a positive correlation between income and vaccination rate. Therefore, in the following part, we would do a test to verify there is a positive correlation between these two factors.

The following is the plot:

Patterns in residual plot suggest that our linear model model may not be appropriate for the data. In this case, you may notice that the residuals corresponding to lower income(under 30,000) tend to be positive, and there seems to be a little bit of clustering of points in the middle. But, overall, the linear model form seems reasonable.

Methods for correlation analyses

Pearson correlation (r), which measures a linear dependence between two variables (x and y). It’s also known as a parametric correlation test because it depends to the distribution of the data. It can be used only when x and y are from normal distribution. The plot of y = f(x) is named the linear regression curve.

## 
##  Pearson's product-moment correlation
## 
## data:  income_edu_v$one_dose_at_least_p and income_edu_v$income
## t = 37.282, df = 12512, p-value < 2.2e-16
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.3003450 0.3318843
## sample estimates:
##      cor 
## 0.316202

The p-value of the test is 2.2e-16, which is less than the significance level alpha = 0.05. We can conclude that wt and mpg are significantly correlated with a correlation coefficient of 0.316202 and p-value of 2.2e-16 .

Summary for this part: There is approximate linear relationship between income and vaccination rate, although the correlation is week.

Normality of Vaccine Distrubution Proportions

As an aside, we can observe that the distributions of the dose distribution proportions across zip codes are approximately normal, giving us greater confidence in our p-values.

Discussion

Broader interpretations of analysis

A possible explanation for the positive correlation for urban environments could stem from people not having to travel as far to receive the vaccine due to the population density being higher.

A possible explanation for the positive correlation for income could stem from people having a better understanding of vaccines and discern misinformation about the inefficacy or risks associated with vaccines.

Potential shortcoming

  • The WI datasets may not be fully representative of all cases nationally or globally
  • Linear models of two variables are simplistic and do not offer the most holistic study review analysis
  • Vaccination rate data is not fully complete and are only point estimates at the time of data retrieval, instead of time-based data

Potential future directions for additional work

  • Since the vaccination process is still in progress, we can wait for a few months and redo the study. Try to prove that what we did is correct.
  • As we only researched in Wisconsin, it might not be true for other places in the world. We can collect more data from other places to verify our conclusion.

Primary conclusions and the primary evidence

From the beginning of the pandemic in early 2020, our lives have been meaningfully interrupted and negatively impacted. With CDC guidance to curb spread by widespread vaccination and ultimately herd immunity, we seek to analyze the effects of demographic factors on populations’ willingness and attitude toward a COVID vaccine. Despite the government providing low or no-cost vaccines in Wisconsin, we still observe some citizens rejecting getting vaccinated because of associated doubts of the effectiveness of and fears associated with a COVID vaccine.

We have found a meaningful positive correlation between higher levels of education and vaccine adoption rates as well as higher income levels and vaccine adoption rates.

Further thoughts & broader implication

The research let us to think that if people are more acknowledged of the benefits of vaccines, would they become more willing to get vaccinated? And will the effectiveness of the vaccine also affect the vaccination rate?

References


  1. Facebook COVID-19 Information Center↩︎

  2. Instagram COVID-19 Information Center↩︎

  3. DHS Vaccine Data↩︎

  4. Homework 6↩︎

  5. ZipCodesToGo↩︎

  6. MADISON.COM↩︎